Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Make late_gc_lowering more robust #57380

Merged
merged 6 commits into from
Feb 17, 2025
+153 −36
Merged

Make late_gc_lowering more robust #57380

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
cedaa66
Make late_gc_lowering more robust to cases where we optimize the sret…
gbaraldi Feb 12, 2025
a34a4a1
Add tests
gbaraldi Feb 13, 2025
52948d3
Remove redundant comments
gbaraldi Feb 13, 2025
91a14c4
Add notsafepoint annotation
gbaraldi Feb 13, 2025
1dbdf34
Fix final GC lowering detection
gbaraldi Feb 14, 2025
3e78f79
Update src/llvm-final-gc-lowering.cpp
gbaraldi Feb 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 32 additions & 3 deletions src/llvm-final-gc-lowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -167,14 +167,14 @@ bool FinalLowerGC::runOnFunction(Function &F)
initAll(*F.getParent());
if (!pgcstack_getter && !adoptthread_func) {
LLVM_DEBUG(dbgs() << "FINAL GC LOWERING: Skipping function " << F.getName() << "\n");
return false;
goto verify_skip;
}

// Look for a call to 'julia.get_pgcstack'.
pgcstack = getPGCstack(F);
if (!pgcstack) {
LLVM_DEBUG(dbgs() << "FINAL GC LOWERING: Skipping function " << F.getName() << " no pgcstack\n");
return false;
goto verify_skip;
}
LLVM_DEBUG(dbgs() << "FINAL GC LOWERING: Processing function " << F.getName() << "\n");
queueRootFunc = getOrDeclare(jl_well_known::GCQueueRoot);
Expand Down Expand Up @@ -212,8 +212,37 @@ bool FinalLowerGC::runOnFunction(Function &F)
#undef LOWER_INTRINSIC
}
}

return true;
// Verify that skipping was in fact correct
verify_skip:
#ifdef JL_VERIFY_PASSES
for (auto &BB : F) {
for (auto &I : make_early_inc_range(BB)) {
auto *CI = dyn_cast<CallInst>(&I);
if (!CI)
continue;

Value *callee = CI->getCalledOperand();
assert(callee);
auto IS_INTRINSIC = [&](auto intrinsic) {
auto intrinsic2 = getOrNull(intrinsic);
if (intrinsic2 == callee) {
errs() << "Final-GC-lowering didn't eliminate all intrinsics'" << F.getName() << "', dumping entire module!\n\n";
errs() << *F.getParent() << "\n";
abort();
}
};
IS_INTRINSIC(jl_intrinsics::newGCFrame);
IS_INTRINSIC(jl_intrinsics::pushGCFrame);
IS_INTRINSIC(jl_intrinsics::popGCFrame);
IS_INTRINSIC(jl_intrinsics::getGCFrameSlot);
IS_INTRINSIC(jl_intrinsics::GCAllocBytes);
IS_INTRINSIC(jl_intrinsics::queueGCRoot);
IS_INTRINSIC(jl_intrinsics::safepoint);
}
}
#endif
return false;
}

PreservedAnalyses FinalLowerGCPass::run(Function &F, FunctionAnalysisManager &AM)
Expand Down
82 changes: 57 additions & 25 deletions src/llvm-late-gc-lowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1114,6 +1114,47 @@
}
}

// Look through instructions to find all possible allocas that might become the sret argument
static SmallSetVector<AllocaInst *, 8> FindSretAllocas(Value* SRetArg) {
SmallSetVector<AllocaInst *, 8> allocas;
if (AllocaInst *OneSRet = dyn_cast<AllocaInst>(SRetArg)) {
allocas.insert(OneSRet); // We found the alloca directly
} else { // Look throught instructions until we find it

Check warning on line 1122 in src/llvm-late-gc-lowering.cpp

View workflow job for this annotation

GitHub Actions / Check for new typos 

perhaps "throught" should be "thought or through or throughout".
SmallSetVector<Value *, 8> worklist;
worklist.insert(SRetArg);
while (!worklist.empty()) {
Value *V = worklist.pop_back_val();
if (AllocaInst *Alloca = dyn_cast<AllocaInst>(V->stripInBoundsOffsets())) {
allocas.insert(Alloca);
} else if (PHINode *Phi = dyn_cast<PHINode>(V)) {
for (Value *Incoming : Phi->incoming_values()) {
worklist.insert(Incoming);
}
} else if (SelectInst *SI = dyn_cast<SelectInst>(SRetArg)) {
auto TrueBranch = SI->getTrueValue();
auto FalseBranch = SI->getFalseValue();
if (TrueBranch && FalseBranch) {
worklist.insert(TrueBranch);
worklist.insert(FalseBranch);
} else {
llvm_dump(SI);
assert(false && "Malformed Select");
}
} else {
llvm_dump(V);
assert(false && "Unexpected SRet argument");
}
}
}
assert(allocas.size() > 0);
assert(std::all_of(allocas.begin(), allocas.end(), [&] (AllocaInst* SRetAlloca){
return (SRetAlloca->getArraySize() == allocas[0]->getArraySize() &&
SRetAlloca->getAllocatedType() == allocas[0]->getAllocatedType());
}
));
return allocas;
}

State LateLowerGCFrame::LocalScan(Function &F) {
State S(F);
SmallVector<int, 8> PHINumbers;
Expand Down Expand Up @@ -1165,46 +1206,35 @@
Type *ElT = getAttributeAtIndex(CI->getAttributes(), 1, Attribute::StructRet).getValueAsType();
auto tracked = CountTrackedPointers(ElT, true);
if (tracked.count) {
AllocaInst *SRet = dyn_cast<AllocaInst>((CI->arg_begin()[0])->stripInBoundsOffsets());
assert(SRet);
{
SmallSetVector<AllocaInst *, 8> allocas = FindSretAllocas((CI->arg_begin()[0])->stripInBoundsOffsets());
// We know that with the right optimizations we can forward a sret directly from an argument
// This hasn't been seen without adding IPO effects to julia functions but it's possible we need to handle that too
// If they are tracked.all we can just pass through but if they have a roots bundle it's possible we need to emit some copies ¯\_(ツ)_/¯
for (AllocaInst *SRet : allocas) {
if (!(SRet->isStaticAlloca() && isa<PointerType>(ElT) && ElT->getPointerAddressSpace() == AddressSpace::Tracked)) {
assert(!tracked.derived);
if (tracked.all) {
S.ArrayAllocas[SRet] = tracked.count * cast<ConstantInt>(SRet->getArraySize())->getZExtValue();
}
else {
Value *arg1 = (CI->arg_begin()[1])->stripInBoundsOffsets();
SmallSetVector<AllocaInst *, 8> gc_allocas = FindSretAllocas(arg1);
AllocaInst *SRet_gc = nullptr;
if (PHINode *Phi = dyn_cast<PHINode>(arg1)) {
for (Value *V : Phi->incoming_values()) {
if (AllocaInst *Alloca = dyn_cast<AllocaInst>(V->stripInBoundsOffsets())) {
if (SRet_gc == nullptr) {
SRet_gc = Alloca;
} else if (SRet_gc == Alloca) {
continue;
} else {
llvm_dump(Alloca);
llvm_dump(SRet_gc);
assert(false && "Allocas in Phi node should match");
}
} else {
llvm_dump(V->stripInBoundsOffsets());
assert(false && "Expected alloca");
}
}
} else {
SRet_gc = dyn_cast<AllocaInst>(arg1);
if (gc_allocas.size() == 1) {
SRet_gc = gc_allocas.pop_back_val();
}
if (!SRet_gc) {
else {
llvm_dump(CI);
llvm_dump(arg1);
assert(false && "Expected alloca");
for (AllocaInst *Alloca : gc_allocas) {
llvm_dump(Alloca);
}
assert(false && "Expected single alloca");
}
Type *ElT = SRet_gc->getAllocatedType();
if (!(SRet_gc->isStaticAlloca() && isa<PointerType>(ElT) && ElT->getPointerAddressSpace() == AddressSpace::Tracked)) {
S.ArrayAllocas[SRet_gc] = tracked.count * cast<ConstantInt>(SRet_gc->getArraySize())->getZExtValue();
}
break; // Found our gc roots
}
}
}
Expand Down Expand Up @@ -1401,6 +1431,8 @@
return S;
}



static Value *ExtractScalar(Value *V, Type *VTy, bool isptr, ArrayRef<unsigned> Idxs, IRBuilder<> &irbuilder) {
Type *T_int32 = Type::getInt32Ty(V->getContext());
if (isptr) {
Expand Down
Loading